Asymmetrical consequences of behavioral change through
reward and punishment
TOBIAS GREITEMEYER1* AND ELLIE KAZEMI2
1
Ludwig-Maximilians-University Munich, Germany
2
California State University, Northridge USA
Abstract
Previous research (Greitemeyer & Weiner, 2003) has demonstrated that compliance, because of an
anticipated reward is attributed more to the person than compliance because of an anticipated
punishment. The present research extended these findings to an educational context. Three studies
revealed that parents who ask their children to change inappropriate behaviors are more likely to
ascribe their children's improvement to the child, if the child was promised a reward, rather than
threatened, to receive a punishment if the child did not improve. Moreover, because a chilďs
improved behavior is more likely to be ascribed to the child given a reward as compared to a
punishment, parents expect that rewards (as opposed to punishments) are more likely to sustain
improved behavior, when the incentive is no longer offered. Finally, participants report to be more
likely to induce behavioral change through reward rather than punishment. Copyright # 2007
John Wiley & Sons, Ltd.
To change an inappropriate behavior, operant conditioning is often applied (Skinner, 1953; Thorndike,
1913). Operant conditioning is defined as a type of learning in which a behavior is strengthened or
weakened, when it is followed by certain consequences. Whereas anticipated rewards (i.e., positive
reinforcers) generally strengthen behavior, punishments generally weaken behavior. Most educators
stress that reward, rather than punishment should be used to change inappropriate behavior (e.g.,
Banks, 2002; Larsen & Tentis, 2003). However, very often no clear rationale for this suggestion has
been given. In the present research, it is examined whether an improved behavior, because of a reward
(as opposed to a punishment) is rather ascribed to the person, and, consequently, rewards more than
punishments are perceived as more likely to sustain improved behavior, even when the incentive is no
longer offered.
Early theorists (Thorndike, 1911) assumed that reward and punishment could be construed as
opposite sides of a coin and symmetrical in their consequences. Presently, however, this assumption has
European Journal of Social Psychology
Eur. J. Soc. Psychol. 38, 246­259 (2008)
Published online 25 January 2007 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/ejsp.418
*Correspondence to: Dr Tobias Greitemeyer, Department of Psychology, University of Munich Leopoldstr. 13, D-80802 Munich,
Germany. E-mail: togre@psy.uni-muenchen.de
Copyright # 2007 John Wiley & Sons, Ltd.
Received 29 June 2006
Accepted 23 November 2006
not remained undisputable. For example, Miller (1944) demonstrated that motivation increases as the
distance to a goal decreases. However, the change in strength of motivation seems to depend on the goal
(i.e., obtaining reward versus avoiding punishment). The slope of motivation for rewards decreases less
as a function of distance from the desired goal than does the strength of motivation to avoid
punishments. Attribution theorists have even contended that reward might decrease, whereas
punishment might increase motivated behavior (Meyer et al., 1979). Persons who are rewarded for
success at an easy task might assume that they have low ability. On the other hand, persons who are
punished for failure might assume that they have the ability to accomplish the task. These beliefs are
associated with enhanced achievement striving (Weiner, 1986). In sum, and opposed to the assumptions
by Thorndike (1913), reward and punishment could be asymmetrical in their consequences on
motivated behavior.
In a different context, recent research has documented that the effect of anticipated rewards versus
punishments on dispositional causation also are not symmetrical. For instance, Alanazi and Rodrigues
(2003), Rodrigues (1995), and Rodrigues and Lloyd (1998) demonstrated that compliance because of
an anticipated reward is attributed more to the person than compliance because of an anticipated
punishment. Assume, for example, a doctor asks a nurse to administer a drug that has not been officially
approved and offers the nurse either a pay raise or threatens that the nurse's wages will be lowered, if
she does not administer the drug. If the nurse complies and administers the drug, the nurse is perceived
more responsible for this transgression, when the compliance follows a promise of a pay raise as
opposed to the threat of a pay cut (see also Reeder & Spores, 1983; Wells, 1980).
In a subsequent series of studies, Greitemeyer and Weiner (2003) showed that the asymmetrical
consequences of reward and punishment on perceptions of responsibility and dispositional attributions
were reliable across a variety of incentives and requested behaviors. More importantly, they
demonstrated that this asymmetry was not based on unequal norms (i.e., differences in perceived
compliance rates), which yielded different dispositional inferences. For example, it was found that if
the doctor offers the nurse a strong incentive (e.g., ``If you do this, I will write a strong letter of
recommendation for the job you applied for next year.'') so that almost all nurses would comply, the
nurse would be perceived as less responsible for complying than if the doctor would have offered a
weak incentive (e.g., ``If you do not do this, I will write a weak letter of recommendation for the job you
applied for next year.'') (Kelley, 1967). In fact, Greitemeyer and Weiner (2003) found that the more
respondents perceived themselves as likely to comply with the request for a transgression, the less the
transgressor was perceived as responsible. However, these authors documented that the asymmetrical
consequences of reward and punishment on perceptions of responsibility remained significant, even
after controlling for perceived compliance rates. Meaning, even though respondents held the
transgressor less responsible if they themselves were likely to comply with the request, when these
compliance rates were controlled for, the asymmetrical consequences of reward and punishment on
perceptions of the transgressor's responsibility still remained significant.
In the present studies, the findings reported by Greitemeyer and Weiner (2003) are applied to an
educational context, in which a series of vignettes depict school-related situations, whereby parents
want their children to change inappropriate behaviors (e.g., disrupting the classroom or being
disrespectful toward the teacher). To motivate their child to improve, the parents either promise a
reward for improvement or threaten a punishment in case the child does not improve. The child
complies with the request; that is, does better. Given positive relative to negative incentives, it is
anticipated that the chilďs improvement should be more ascribed to the child. In addition to applying
the previous research by Greitemeyer and Weiner (2003) to a new context, in Study 1 and Study 2, this
work was applied to a sample of parents. The present research also goes beyond the previous work by
examining whether differences in perceived dispositional causation affected perceptions of long-term
effects of positive and negative incentives.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 247
EFFECTS OF POSITIVE AND NEGATIVE INCENTIVES ON LEARNING
Rewarding students for learning has often been criticized to undermine intrinsic motivation (e.g., Deci,
Koestner, & Ryan, 1999). Because a reward for a behavior may result in the student ascribing a
successful performance to the reward, rather than to the intrinsic positive properties of the activity,
motivated performance may decrease (Lepper, Greene, & Nisbett, 1973; Lepper, Keavney, & Drake,
1996). Such motivation researchers contend that the usage of reinforcement strategies decreases
individuals' perceptions of competence and self-determination, thereby decreasing individuaľs
intrinsic motivation to master or perform a task. Therefore, motivation researchers, concerned with the
rise of usage of reward contingency systems by educators have gone as far as referring to the usage of
external rewards and even verbal praise as bribery (Kohn, 1993).
However, other researchers have argued that rewards increase intrinsic motivation, when used
appropriately (Eisenberger, Pierce, & Cameron, 1999). When rewards provide a student with
information about their mastery of a task, or when the rewards are offered contingent on performance,
then improvement should continue, even when the reward is no longer offered. The debate, regarding
the usage of external reinforces on intrinsic motivation is inconclusive and two recent meta-analytic
studies yielded contradictory results (Cameron, Banko, & Pierce, 2001; Deci et al., 1999). In contrast,
most educators and psychologists agree that punishment does not lead to any positive behavior and,
thus, tends to only to suppress behavior. Once the threat of punishment is taken away, the child may
continue to use the undesirable behavior. Thus, it is expected that parents anticipate rewards more than
punishments to sustain improved behavior when the incentive is no longer offered. Finally, this effect is
expected to be mediated by ascriptions of dispositional causation: Because an improved behavior is
rather ascribed to the person given rewards rather than punishments, rewards are more likely to sustain
improved behavior even when the reward is no longer offered. This research question has not yet been
examined.
To sum up, the following hypotheses were tested in the present research:
Hypothesis 1: There are greater dispositional attributions to the child for behavioral change given a
reward than a threatened punishment.
Hypothesis 2: Given a reward, relative to a punishment, participants expect that the improved
behavior will be sustained even when the incentive is no longer offered.
Hypothesis 3: The effect of Hypothesis 2 is mediated by the effect of Hypothesis 1: Inasmuch as
participants rather ascribe the chilďs improved behavior to the child given a reward as compared to a
punishment, they expect that rewards have more long-lasting outcomes.
To test these hypotheses, three studies were conducted. In a first study, parents were told to imagine
that their child has poor grades in school. They were asked to imagine that they use different positive
and negative incentives to change their chilďs low school achievement, and the parent was told that
the child did, in fact, improve. For each incentive, participants indicated to what extent the
improvement was due to something about the child and whether they expected the child to continue to
improve, even when the incentive was no longer offered. In addition, perceived behavior rates were
assessed. This allowed for covariation of these judgments from the dispositional causation ratings, and
thus to rule out that the greater attributions to the person given rewards rather than punishments could
be ascribed to unequal behavior rates. The second study aimed to replicate the first study, with parents
imagining that their child behaved badly in school rather than had poor grades. In the third study, it was
further tested whether participants, because of their expectation that reward more than punishment will
sustain improved behavior, when the incentive is no longer offered, were more likely to report to induce
behavioral change through reward rather than punishment.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
248 Tobias Greitemeyer and Ellie Kazemi
STUDY 1
Participants were asked to imagine that they used different positive and negative incentives to change
their children's low school achievement. On an explorative basis, we examined whether different
causes of the chilďs poor performance moderated the effect of incentive valence on parents' reactions
(although this was not expected). Previous work has documented that effort and ability are perceived as
the main causes determining achievement (Weiner, 1986). Hence, whether the chilďs performance was
due to low effort or despite high effort, and whether the chilďs performance was due to low ability or
despite high ability, were manipulated in a factorial design.
Method
Participants were 121 adults (92 women and 29 men) who had at least one child (M  2.16, SD  0.86,
range  1­6). The participants' ages ranged between 28 and 63 (M  39.21, SD  6.05). All
participants were approached in different public buildings in Munich/Germany by either a female or a
male research assistant and were asked to fill out the questionnaire. At the onset of the questionnaire,
participants were asked to imagine that their chilďs school performance was poor. In the condition low
effort/low ability, participants read:
``You receive a phone call from school. A teacher informs you that your child has very bad grades in
all tests. According to the teacher, this is due to the fact that your child is lazy. The child does not do the
homework, tends to skip class, does not pay attention, and is not participating in the lessons. In addition,
your chilďs abilities are not sufficient enough to handle the class material and your child has
concentration problems. You are very worried about your chilďs poor academic performance and want
your child to do better. What incentive you use is given below.''
In the high effort conditions, it was said that the child has very bad grades in all tests, in spite of the fact
that the child tries hard to be successful in school. The child turns in the homework on time, is present every
day, pays attention, and participates actively in the lessons. In the high ability conditions, it was said that
the child has very bad grades in all tests, in spite of the fact that the child is smart and concentrated.
Participants then read 32 different incentives or behaviors. Sixteen of the incentives/behaviors were
positive and 16 were negative. A positive incentive/behavior was operationalized as a reward for doing
better. A negative incentive/behavior indicated a threat for not doing better. The content of the positive
incentives/behaviors was matched by a corresponding negative incentive/behavior (see Table 1). An
example of a positive incentive was ``If you do better, I will allow you stay up longer during the
weekends,'' and the matched negative incentive, ``If you do not do better, I will make you go to bed
earlier during the weekends.'' A positive behavior was ``You express more affection,'' with the negative
behavior ``You withhold affection.'' In all of the present studies, the effects of incentive and behavior on
the main dependent variables were similar. For readability, incentives and behaviors are referred to as
incentives. Each participant responded to all 32 incentives. Hence, a 2 (Incentive Valence) Â 16 (Kind
of Incentive) Â 2 (Effort: high vs. low) Â 2 (Ability: high vs. low) factorial design, with repeated
measures on the first two factors was employed.
For each incentive, participants indicated the percentage of children that they believed would do
better, given the incentive (behavior rates). Then, participants learned that the child does in fact
improve. That is, participants read that the child has better grades. For each incentive, participants then
responded to one question measuring dispositional causation (on a scale from 1 not at all to 9 very
much). The corresponding question was ``To what extent is the chilďs improvement due to something
about the child (personality, traits, etc.)?''. Participants also indicated the long-term effects of the
incentives. They were asked how the child would do when the incentive is no longer offered (on a scale
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 249
Table 1. Mean ratings for behavior change, dispositional attributions, and continued improvement as a function
of kinds of positive and negative incentive (Studies 1 and 2)
Positive negative incentive
Study 1 Study 2
B D I B D I
Incentive
I will give you more pocket money every month 43.2 6.00 0.54 33.1 5.57 0.56
I will lessen your pocket money every month 39.9 5.13 0.08 27.1 5.14 0.07
I will allow you to watch more TV 37.8 5.83 0.17 25.4 5.30 0.51
I will make you watch less TV 40.8 4.88 0.17 24.6 4.85 0.05
I will allow you to spend more time with your friends 36.1 5.96 0.92 26.2 5.55 0.61
I will make you spend less time with your friends 34.5 5.25 0.29 23.4 4.99 0.01
I will do more activities with you that you like 35.2 6.08 1.08 26.9 5.55 0.54
I will do fewer activities with you that you like 32.6 5.38 0.29 20.1 4.78 0.21
I will take you to the special events you always
wanted to go to
33.8 4.83 0.25 31.8 5.15 0.31
I will not take you to the special events you always
wanted to go to
33.8 5.13 0.04 20.8 4.74 0.13
I will allow you to stay more nights over at your
frienďs house
35.7 5.58 0.33 25.6 5.36 0.46
I would not allow you to stay as many nights over
at your frienďs house as you do now
31.1 5.00 0.13 20.2 4.97 0.04
I will allow you to stay up longer during the weekends 29.7 5.25 0.71 21.6 5.31 0.57
I will make you go to bed earlier during the weekends 35.5 5.42 0.13 19.7 4.89 0.04
I will allow you to go to the sport training more often
than you are going now
28.1 5.75 0.67 23.5 5.10 0.50
I will make you to go to the sport training less
often than you are going now
32.0 5.29 0.08 20.7 4.83 0.02
I will buy you more season tickets for an event
of your choice
29.7 5.08 0.13 25.9 5.15 0.19
I will stop buying you season tickets for an event
of your choice
25.5 5.00 0.13 17.9 4.60 0.08
I will take you out to dinner at your favorite
restaurant more regularly
25.5 5.00 0.17 17.1 4.76 0.16
I will not take you out for dinner at your favorite
restaurant more regularly
24.3 4.71 0.38 11.6 4.50 0.08
I will take you shopping regularly 20.9 4.21 0.04 18.3 4.73 0.16
I will not take you shopping regularly 13.3 3.92 0.17 12.2 4.51 0.08
Behavior pattern
You express more affection 44.0 6.92 1.58 38.5 6.80 1.17
You withhold affection 25.5 4.33 0.17 12.1 4.39 0.14
You praise the child more 42.9 6.96 2.00 40.9 6.79 0.93
You do not praise the child as much 22.5 4.08 0.54 11.7 4.49 0.28
You express your disappointment less 20.1 5.29 0.29 25.9 5.66 0.47
You express your disappointment more 42.5 5.54 0.83 22.7 4.97 0.18
You tell your child more often how you appreciate her 37.2 6.63 1.67 38.4 6.56 1.02
You tell your child less often how you appreciate her 23.4 4.75 0.04 14.0 4.27 0.46
You express less anger 22.3 4.71 0.08 25.1 5.45 0.53
You express more anger 32.5 4.58 0.21 15.2 4.41 0.24
Note: The positive incentive followed the words: ``If you do better,'' the negative incentive followed the words: ``If you do not do
better.''
B, behavior change; D, dispositional attribution; I, continued improvement.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
250 Tobias Greitemeyer and Ellie Kazemi
from 4 will do worse to 4 will do better). Each incentive was immediately followed by these two
items. Question order was counterbalanced. Question order as well as incentive order had no effect on
any of the main dependent variables in any of the following experiments. Thus, these variables are not
considered further. As manipulation checks, participants indicated to what extent is their chilďs low
school achievement due to low effort and to what extent is their chilďs low school achievement due to
low ability, respectively (on scales from 1 not at all to 9 very much).
Results and Discussion
Manipulation Checks
Parents were more likely to agree that their children's low school achievement was due to low effort in the
low effort conditions (M  5.33, SD  2.00) than in the high effort conditions (M  4.31, SD  2.36),
t(111)  2.47, p < .05. Parents were also more likely to agree that their children's low school achievement
was due to low ability in the low ability conditions (M  3.96, SD  2.37) than in the high ability
conditions (M  2.64, SD  1.51), t(109)  3.52, p < .01. Thus, the manipulations were successful.
Behavior Rates
A 2 (Incentive Valence) Â 16 (Kind of Incentive) Â 2 (Effort) Â 2 (Ability) ANOVA with repeated
measures on the first two factors revealed a significant main effect for Kind of Incentive, F(15,
1485)  12.61, p < .001, h2
 .11, indicating that the incentives differed in the perceived likelihood of
improving the chilďs behavior (see Table 1). In addition, there was a main effect of Incentive Valence,
F(1, 99)  90.09, p < .001, h2
 .48. The perceived likelihood of the chilďs improved behavior was
higher for the positive (M  27.8%) than for the negative (M  18.4%) incentives. The main effects of
Effort and Ability, respectively, as well as the interactions between Incentive Valence and Effort and
between Incentive Valence and Ability, respectively, were not significant.
Dispositional Causation
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 101)  45.34, p< .001, h2
 .31.
There were higher ratings of dispositional causation for the positive (M  5.55) than for the negative
incentives (M  4.71). Thus, Hypothesis 1 was confirmed. The main effects of Effort and Ability,
respectively, as well as the interactions between Incentive Valence and Effort and between Incentive
Valence and Ability, respectively, were not significant. A 2 (Incentive Valence)Â 16 (Kind of
Incentive) Â 2 (Effort) Â 2 (Ability) ANCOVA, with repeated measures on both factors, using the behavior
rates for each incentive as covariates, was performed next. Most importantly, the main effect for Incentive
Valence remained significant after controlling for behavior rates, F(1, 85) 9.89, p< .01, h2
 .10.
Effect of Incentive When no Longer Offered
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 98)  37.32, p < .001,
h2
 .28. Parents expected the child would continue to improve, given a positive (M  0.54) rather
than a negative incentive (M  0.07). Moreover, only the positive incentives yielded an expectation of
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 251
continued improvement, t(148)  5.07, p < .001 (for the negative incentives, t(145)  0.30, p  .77).
Thus, Hypothesis 2 received support from the data. The main effects of Effort and Ability, respectively,
as well as the interactions between Incentive Valence and Effort and between Incentive Valence and
Ability, respectively, were not significant.
To test whether perceived dispositional causation mediated the effect of Incentive Valence on
continued improvement, a 2 (Incentive Valence) Â 16 (Kind of Incentive) 2 (Effort) Â 2 (Ability)
ANCOVA, with repeated measures on the first two factors, using perceived dispositional causation for
each incentive as covariates, was performed on the data. The main effect for Incentive Valence was no
longer significant when controlling for dispositional causation, F(1, 96)  1.28, p  .26, h2
 .01. In
contrast, the effect of the covariate was significant, F(1, 96)  7.52, p < .01, h2
 .07. Thus, as
suggested by Hypothesis 3, perceptions of dispositional causation mediated the effect of Incentive
Valence on continued improvement. That is, participants expected that rewards have more long-lasting
outcomes than punishments, because they were more likely to ascribe the chilďs improved behavior to
the child given a reward rather than a punishment.
In sum, given rewards rather than punishments, participants were more likely to ascribe a chilďs
improvement to the child. In addition, this effect remained significant after controlling for behavior
rates. Thus, the asymmetrical effects of reward and punishment on dispositional attribution cannot be
ascribed to differences in social norms. Participants also expected that the improved behavior would be
more likely to sustain when the reward, relative to the punishment, is no longer offered. Moreover,
perceptions of dispositional causation mediated the effect of incentive valence on continued
improvement. That is, participants expected that rewards have more long-lasting outcomes than
punishments, because they ascribe the chilďs improved behavior to the child. The aim of Study 2 was
to replicate the pattern of results reported above. Additionally, participants were asked about poor
behavior rather than poor academic achievement. The design and procedure of Study 2 were very
similar to that of Study 1. However, a different vignette was used.
STUDY 2
Method
Participants were 32 German adults (26 women and 6 men) who had at least one child (M  2.34,
SD  0.83, range  1­5). The participants' ages ranged between 31 and 47 (M  39.26, SD  4.75).
The same incentives and the same dependent measures (besides the manipulation checks) were used as
in Study 1. However, participants were asked to imagine that their child behaved badly in school.
Concretely, they read:
``You receive a phone call from school. A teacher informs you that your child is being disrespectful to
the teachers, is not waiting in line for his/her turn, and is aggressive to classmates. You are worried about
your chilďs bad behavior in school and ask your child to do better. What incentive you use is given below.''
Results and Discussion
Behavior Rates
A 2 (Incentive Valence) Â 16 (Kind of Incentive) ANOVA with repeated measures on both factors
revealed a significant main effect for Kind of Incentive, F(15, 375)  5.40, p < .001, h2
 .18,
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
252 Tobias Greitemeyer and Ellie Kazemi
indicating that the incentives differed in the perceived likelihood of improving the chilďs behavior (see
Table 1). The main effect of Incentive Valence was not significant, F(1, 25)  2.43, p  .13, h2
 .09.
The perceived likelihood of improving the chilďs behavior was relatively equal for the positive
(M  32.6%) and the negative (M  30.6%) incentives.
Dispositional Causation
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 23)  10.15, p < .01,
h2
 .31. As in Experiment 1, there were higher ratings of dispositional causation for the positive
(M  5.63) than for the negative incentives (M  4.90).
Effect of Incentive When no Longer Offered
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 23)  11.17, p < .01,
h2
 .33. Participants expected that the child would continue to improve, given a positive (M  0.66)
rather than a negative incentive (M  0.05). Moreover, compared to the scale midpoint, only the
positive incentives yielded an expectation of continued improvement, t(28)  1.96, p  .06 (for the
negative incentives, t(27)  0.01, p  .99). Moreover, the main effect for Incentive Valence was no
longer significant when controlling for dispositional causation, F(1, 22)  2.64, p  .12, h2
 .11.
However, in contrast to Study 1, the effect of the covariate was not significant, F(1, 22)  0.47, p  .50,
h2
 .02.
In sum, as in Study 1 (while using a different vignette), given rewards rather than punishments,
participants were more likely to ascribe a chilďs improvement to the child, and, thus, they expected that
the improved behavior would be more likely to sustain when the incentive is no longer offered. Note,
however, that the mediator (dispositional causation) was not positively related with the outcome (effect
of incentive when no longer offered) while controlling for the predictor variable (incentive valence).
This (nonsignificant) finding could be simply due to the small sample size in Study 2. However, to
clarify whether ratings of long-term effects of incentive indeed are positively affected by ratings of
dispositional causation, a third study was conducted.
STUDY 3
In Study 3, we also tested whether participants were more likely to report to induce behavioral change
through reward rather than punishment. Educators, as well as lay people are convinced that reward is
better than punishment for learning (e.g., Banks, 2002; Larsen & Tentis, 2003). However, although
educators invariably advocated reward rather than punishment, no clear rationale has been given.
Therefore, we examined why rewards are more likely to be offered than punishments. In Studies 1 and 2,
given rewards rather than punishments, participants expected that the improved behavior would be
more likely to sustain when the incentive is no longer offered. Because of the expectation that reward,
more than punishment will sustain improved behavior when the incentive is no longer offered, it is
anticipated that participants report to be more likely to offer rewards rather than punishments. To test
this reasoning, participants also indicated the likelihood that they offer their child a positive or negative
incentive, respectively, to change their chilďs behavior.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 253
Method
Participants were 65 students (44 women and 21 men) of the Ludwig-Maximilians-University Munich
who participated for course credit. The participants' ages ranged between 19 and 40 (M  24.92,
SD  4.80). The great majority of the participants (N  63) did not have children. One participant had
one child and another participant had three children. Participants received either the low achievement
vignette or the bad behavior vignette. The same vignettes were used as in the previous studies.
Participants responded to eight positive and eight negative incentives (see Table 2). A 2 (Incentive
Valence) Â 8 (Kind of Incentive/Behavior) Â 2 (Type of Behavior: bad behavior vs. low achievement)
factorial design, with repeated measures on the first two factors was employed. The same dependent
variables were measured as in Experiments 1 and 2. In addition, participants responded to two items
measuring how likely it is that they would offer their child a positive (negative) incentive to change the
chilďs behavior (on a scale from 1 not likely to 9 very likely).
Table 2. Mean ratings for behavior change, dispositional attributions, continued improvement, and likelihood of
offering incentive, as a function of kinds of positive and negative incentive and kind of behavior (Study 3)
positive negative incentive
Bad behavior Low achievement
B D I B D I
Incentive
I will give you more pocket money every month 64.6 3.77 0.67 43.8 5.28 0.03
I will lessen your pocket money every month 54.5 3.16 1.20 32.2 4.88 0.53
I will allow you to spend more time with
your friends
51.7 5.48 0.47 40.9 5.84 0.17
I will make you spend less time with
your friends
50.3 5.10 0.63 34.0 5.31 0.50
I will do more activities with you that you like 47.2 6.10 0.90 33.9 5.84 0.20
I will do fewer activities with you that you like 36.6 5.32 0.40 24.3 5.03 0.17
I will allow you to stay more nights over at
your frienďs house
48.9 5.26 0.67 34.7 5.50 0.03
I would not allow you to stay as many nights over at
your frienďs house as you do now
41.5 4.61 1.00 29.5 5.38 0.30
I will allow you to stay up longer than normally
during the weekends
38.2 5.16 0.03 26.8 5.44 0.27
I will make you go to bed earlier than normally
during the weekends
39.7 4.16 1.03 23.7 5.00 0.23
I will allow you to go to the sport training more often
than you are going now
36.7 5.68 0.47 32.1 5.66 0.10
I will make you to go to the sport training less often
than you are going now
38.2 4.58 0.53 24.8 5.13 0.17
Behavior pattern
You express more affection 53.4 6.77 1.17 43.4 6.31 0.23
You withhold affection 42.2 4.32 0.53 21.0 5.28 0.90
You praise the child more 48.8 6.52 0.97 44.9 6.38 0.80
You do not praise the child as much 23.7 4.68 0.10 16.9 5.66 0.03
Note: The positive incentive followed the words: ``If you do better,'' the negative incentive followed the words: ``If you do not do
better.''
B, behavior change; D, dispositional attribution; I, continued improvement.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
254 Tobias Greitemeyer and Ellie Kazemi
Results and Discussion
Behavior Rates
A 2 (Incentive Valence) Â 8 (Kind of Incentive) Â 2 (Type of Behavior) ANOVA with repeated
measures on the first two factors revealed a significant main effect for Kind of Incentive, F(7,
434)  12.41, p < .001, h2
 .17, indicating that the incentives differed in the perceived likelihood of
improving the chilďs behavior (see Table 2). In addition, the main effect of Type of Behavior was also
significant, F(1, 62)  8.58, p < .01, h2
 .12. The perceived likelihood that the chilďs behavior
improves was higher for bad behavior (M  44.8%) than for low achievement (M  32.5%). Finally,
there was a main effect of Incentive Valence, F(1, 62)  47.27, p < .001, h2
 .43. The perceived
likelihood of improving the chilďs behavior was higher for the positive (M  43.4%) than for the
negative (M  33.8%) incentives.
Dispositional Causation
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 61)  46.48, p < .001,
h2
 .43. There were higher ratings of dispositional causation for the positive (M  5.69) than for the
negative incentives (M  4.85). Moreover, in an ANCOVA, using behavior rates for each incentive as
covariates, the main effect for Incentive Valence remained significant after controlling for behavior
rates, F(1, 59)  23.41, p < .001, h2
 .28. In addition, the ANOVA revealed a significant interaction
between Incentive Valence and Type of Behavior, F(1, 61)  4.59, p < .05, h2
 .07. Given bad
behavior, there were higher ratings of dispositional causation for the positive (M  5.59) than for the
negative incentives (M  4.49), F(1, 30)  34.14, p < .001, h2
 .53. Similarly, given low school
achievement, there were higher ratings of dispositional causation for the positive (M  5.78) than for
the negative incentives (M  5.21), F(1, 31)  13.10, p < .01, h2
 .30. However, the effect was not as
pronounced.
Effect of Incentive When no Longer Offered
An ANOVA revealed a significant main effect for Incentive Valence, F(1, 59)  25.68, p < .001,
h2
 .30. Participants expected the child would continue to improve, given a positive (M  0.36)
rather than a negative incentive (M  0.30). Compared to the scale midpoint, the effect for the
positive incentives, t(63)  2.36, p < .05 was significant; the effect for the negative incentives,
t(62)  1.77, p  .08, was marginally significant. In addition, the interaction between Incentive
Valence and Behavior was significant, F(1, 59)  4.53, p < .05, h2
 .07. Given bad behavior,
participants expected that the child would continue to improve, given a positive (M  0.49) rather
than a negative incentive (M  0.45), F(1, 29)  23.24, p < .001, h2
 .45. Similarly, given low
school achievement, participants also expected that the child would continue to improve, given a
positive (M  0.23) rather than a negative incentive (M  0.16), F(1, 30)  4.84, p < .05, h2
 .14.
However, the effect was not as pronounced.
To test whether perceived dispositional causation mediated the effect of Incentive Valence on
continued improvement, a 2 (Incentive Valence) Â 8 (Kind of Incentive) 2 (Behavior) ANCOVA, with
repeated measures on the first two factors, using perceived dispositional causation for each incentive as
covariates, was performed on the data. The main effect for Incentive Valence, F(1, 58)  1.87, p  .18,
h2
 .03, was no longer significant, when controlling for dispositional causation. In contrast, the effect
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 255
of the covariate was significant, F(1, 58)  38.39, p < .001, h2
 .40. Thus, perceptions of dispositional
causation appeared to mediate the effect of Incentive Valence on continued improvement.
Likelihood of Offering Incentive
Participants would rather offer a positive (M  6.35) than a negative incentive (M  4.63), t(64)  4.68,
p < .001. However, the difference between likelihood of offering positive and negative incentives was
not significantly associated with the difference between ratings of effect of incentive when no longer
offered for the mean of the positive and the negative incentives, r(65)  .11, p  .37.
GENERAL DISCUSSION
The present research replicated and extended previous research (Greitemeyer & Weiner, 2003, 2006)
on the asymmetrical effects of reward and punishment on dispositional attribution. Replicating the
findings from previous studies, there were greater attributions to the person given rewards rather than
punishments. Specifically, participants were more likely to ascribe a chilďs improvement to the child in
situations of reward relative to punishment. Moreover, the asymmetrical effects of reward and
punishment on dispositional attributions were not due to differences in perceived behavior rates.
Although two of the experiments revealed significant differences between the positive and negative
incentives employed in the present studies in terms of behavior rates, the effect of incentive valence on
dispositional attributions remained significant when statistically controlling for behavior rates. Thus,
the asymmetrical effects of reward and punishment on dispositional attribution cannot be ascribed to
differences in social norms. Extending previous research, given reward rather than punishment,
participants were more likely to expect that the improved behavior would sustain, when the incentive
was no longer offered. This finding was mediated by perceptions of dispositional causation: Because
participants were more likely to ascribe the chilďs improved behavior to the child given reward rather
than punishment, they expected that rewards have more long-lasting outcomes than punishments. This
pattern of results occurred in two of the three studies, using different vignettes and employing a college
sample as well as real parents. In Study 2, the effect of incentive valence on continued improvement
also disappeared when controlling for dispositional causation. However, dispositional causation was
not significantly associated with continued improvement, which could be due to small sample size.
Study 3 further revealed that participants were more likely to report to induce behavioral change
through reward rather than punishment. In line with this finding, educators stress the positive effects of
reward relative to punishment for learning (e.g., Banks, 2002; Larsen & Tentis, 2003). In addition,
various side effects of punishment have been stressed, such as anger or feelings of hostility, which could
also interfere with learning. However, several studies have revealed that learning is more strongly
affected by punishment than by reward (for a review, see Baumeister, Bratlavsky, Finkenauer, & Vohs,
2001). For example, in a study by Costantini and Hoving (1973), children were either instructed to win
as many marbles as possible or to lose as few marbles as possible. Results showed that avoiding losing
marbles resulted in more effective inhibition of responses than trying to gain marbles. Tindall and
Ratliff (1974) also compared the effects of reward and punishment on performance. In the reward
condition, children (either first, fourth, or eighth graders) received a token for correctly identifying
figures. In the punishment condition, children were exposed to a loud noise for incorrect responses.
Children performed significantly better in the punishment condition than in the reward condition. In
addition, developmental stage did not show any interaction with experimental condition.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
256 Tobias Greitemeyer and Ellie Kazemi
Interestingly, recent research (Kazemi & Weiner, in preparation) revealed that parents punish more
than they reward. In their study, Southern Californian parents of school-aged children were asked what
they did to get their child to do better, when they had done poorly in school or when they had behaved
badly in school. Parents reported offering punishment about three times more often than offering
reward. Thus, it seems that, although parents' explicit attitudes toward punishment are negative, their
behavior is in line with the above-described research showing that punishment is better than reward for
learning. Research by Cohen, Manimala, and Blount (2000) also showed that parents' explicit attitudes
and behaviors do not correspond. In their study, parents were asked to provide reports of the therapeutic
behaviors they typically engage in during their children's immunizations. Then, the immunization
procedure was video taped and coded. Results indicated that parents overestimated the quantity of their
therapeutic behaviors and that no relationship existed between parents' reports of their behavior and
their actual behavior during children's immunization.
But why do parents' explicit attitudes and behaviors not correspond? It is conceivable that people
may simply lack self knowledge (Wilson & Dunn, 2004). For instance, parents might be motivated to
keep their beliefs that punishment is more effective than reward outside of unconsciousness. Moreover,
previous research has shown that, when people's explicit and implicit attitudes dissociate, their
behavior is rather determined by their implicit than their explicit attitudes (Wilson, Lindsey, &
Schooler, 2000). Thus, whereas parents' explicit attitudes toward punishment are negative, their
implicit attitudes might be positive. Furthermore, a parent whose child behaves badly is likely to
experience anger, which in turn is associated with aggression (Weiner, 1995). Thus, even if both
explicit and implicit attitudes towards reward are more positive than towards punishment, parents
might impulsively punish rather than reward. Finally, and in contrast to our expectations, the greater
likelihood to induce behavioral change through rewards, rather than punishments in Study 3 was not
associated with perceptions of continued improvement. That is, parents do not prefer to reward than
punish, because they are convinced that reward has more positive long-lasting effects than punishment.
It is, thus, conceivable that parents reported preferring to reward because they have been told by
educators that reward is better than punishment, but parents may not be sure as to why they reward
versus punish. In addition, parents may simply indicate to rather use rewards than punishments due to
social desirability concerns. Hence, because their favorable attitudes toward rewarding are not
grounded in strong beliefs (Petty & Krosnick, 1995), they may not practice what they preach. Future
research is needed to examine these issues and other explanations for the apparent discrepancy between
parents' attitudes and behaviors.
Limitations
The reader should be aware that in all three studies, a within-participants design was employed. That is,
participants responded to all positive as well to all negative incentives. Thus, participants may have
compared the different kinds and valence of incentives, within each study and calibrated their
judgments to the incentive manipulations. Note also that participants' reported likelihood of offering
incentive was not assessed spontaneously, but after various relevant questions that may have affected
participants' responding. It is also worth noting that participants received only limited information
about the circumstances of the inappropriate behavior of the child. Moreover, because all vignettes
were hypothetical, no judgments were made by participants that involved real consequences. Hence,
future research in which actual behavior is assessed would be fruitful. For instance, parents' responses
to their children's behavior could be coded in terms of positive and negative incentives used, and their
corresponding dispositional attributions for their children's progress examined. It is also worth noting
that in all of the present studies, parents were asked how they would motivate their children to improve.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 257
It would be interesting to investigate whether individuals also rather induce behavioral change through
reward than punishment in other relationships. Note also that Study 3 contained only a small minority
of participants who were parents. Thus, most of these participants have no experience in raising
children. In the present research, participants were asked to imagine that either one positive or negative
incentive was used to change a chilďs inappropriate behavior. Note, however, that most parents rather
practice trial and error and do not use systematically reward and punishment. In addition, they almost
always use more than one single technique to change a chilďs behavior. Finally, all participants of the
present studies were from Germany, which is a rather liberal country. It is conceivable that people from
more traditional cultures have more favorable attitudes towards punishment.
With these limitations in mind, one may still conclude that most of the findings from the three
studies were consistent and robust. In all studies, given reward relative to punishment, participants were
more likely to ascribe a chilďs improvement to the child. As a consequence, they were more likely to
expect that the improved behavior would sustain when the incentive was no longer offered, which is in
line with educators' plea to use reward rather than punishment (Banks, 2002; Larsen & Tentis, 2003).
However, whether it is not only assumed that reward has more positive long-term effects than
punishment, but reward, in fact, is more likely to sustain improved behavior awaits further empirical
evidence.
ACKNOWLEDGEMENTS
The research reported in this article was made possible by German Scientific Foundation Grant GR
1882/5-1. We are grateful to Bernard Weiner for his helpful comments on the manuscript.
REFERENCES
Alanazi, F. M., & Rodrigues, A. (2003). Power bases and attribution in three cultures. Journal of Social
Psychology, 143, 375­395.
Banks, J. A. (2002). An introduction to multicultural education (3rd ed.). Boston: Allyn and Bacon.
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of
General Psychology, 5, 323­370.
Cameron, J., Banko, K. M., & Pierce, W. D. (2001). Pervasive negative effects of rewards on intrinsic motivation:
The myth continues. The Behavior Analyst, 24, 1­44.
Cohen, L. L., Manimala, R., & Blount, R. L. (2000). Easier said than done: What parents say they do and what they
do during children's immunization. Children's Health Care, 29, 79­86.
Costantini, A. F., & Hoving, K. L. (1973). The effectiveness of reward and punishment contingencies on response
inhibition. Journal of Experimental Child Psychology, 16, 484­494.
Deci, E., Koestner, R., & Ryan, R. M. (1999). A meta-analytic review of experiments examining the effects of
extrinsic rewards on intrinsic motivation. Psychological Bulletin, 125, 627­668.
Eisenberger, R., Pierce, W. D., & Cameron, J. (1999). Effects of reward on intrinsic motivation--Negative, neutral,
and positive: Comment on Deci, Koestner, and Ryan (1999). Psychological Bulletin, 125, 677­691.
Greitemeyer, T., & Weiner, B. (2003). The asymmetrical consequences of reward and punishment on attributional
judgments. Personality and Social Psychology Bulletin, 29, 1371­1382.
Greitemeyer, T., & Weiner, B. (2006). Effects of reward and coercion for transgression compliance on judgments of
responsibility and recommended punishment. Journal of Applied Social Psychology, 36, 3015­3030.
Kazemi, E., & Weiner, B. (in preparation). Reported parental disciplinary tactics to enhance academic and
behavioral performance: An attributional approach. Manuscript being prepared for publication.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
258 Tobias Greitemeyer and Ellie Kazemi
Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium on
motivation (pp. 192­238). Lincoln: University of Nebraska Press.
Kohn, A. (1993). Punished by rewards: The trouble with gold stars, incentive plans, A's, praise and other bribes.
Boston: Houghton Mifflin.
Larsen, M. A., & Tentis, E. (2003). The art and science of disciplining children. Pediatric Clinics of North America,
50, 817­840.
Lepper, M. R., Greene, D., & Nisbett, R. E. (1973). Undermining children's intrinsic interest with extrinsic reward:
A test of the ``overjustification'' hypothesis. Journal of Personality and Social Psychology, 28, 129­137.
Lepper, M. R., Keavney, M., & Drake, M. (1996). Intrinsic motivation and extrinsic rewards: A commentary on
Cameron and Pierce's meta-analysis. Review of Educational Research, 66, 5­32.
Meyer, W. U., Bachmann, M., Biermann, U., Hempelmann, M., Ploeger, F. O., & Spiller, H. (1979). The
informational value of evaluative behavior: Influences of praise and blame on perceptions of ability. Journal of
Educational Psychology, 71, 259­268.
Miller, N. E. (1944). Experimental studies in conflict. In J. McV. Hunt (Ed.), Personality and the behavior disorders
(Vol. 1, pp. 431­465). New York: Ronald Press.
Petty, R. T., & Krosnick J. A. (Eds.) (1995). Attitude strength: Antecedents and consequences. Mahwah, NJ:
Erlbaum.
Reeder, G. L., & Spores, J. M. (1983). The attribution of morality. Journal of Personality and Social Psychology,
44, 736­745.
Rodrigues, A. (1995). Attribution and social influence. Journal of Applied Social Psychology, 25, 1567­1577.
Rodrigues, A., & Lloyd, K. L. (1998). Reexamining bases of power from an attributional perspective. Journal of
Applied Social Psychology, 28, 973­997.
Skinner, B. F. (1953). Science and Human Behavior. New York: Macmillan.
Thorndike, E. L. (1911). Animal intelligence. New York: Macmillan.
Thorndike, E. L. (1913). Educational psychology: The psychology of learning. New York: Teachers College Press.
Tindall, R. C., & Ratliff, R. G. (1974). Interaction of reinforcement conditions and developmental level in a
two-choice discrimination task with children. Journal of Experimental Child Psychology, 18, 183­189.
Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer Verlag.
Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Guilford
Press.
Wells, G. (1980). Asymmetric attributions for compliance: Reward vs. punishment. Journal of Experimental Social
Psychology, 16, 47­60.
Wilson, T. D., & Dunn, E. (2004). Self-knowledge: Its limits, value, and potential for improvement. Annual Review
of Psychology, 55, 493­518.
Wilson, T. D., Lindsey, S., & Schooler, T. (2000). A model of dual attitudes. Psychological Review, 107, 101­126.
Copyright # 2007 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 38, 246­259 (2008)
DOI: 10.1002/ejsp
Asymmetrical consequences 259